The effect of absorbent grid preparation method on precision and accuracy
of ambient nitrogen dioxide measurements using Palmes passive diffusion
tubes
Mathew R. Heal*
School of Chemistry, University of Edinburgh, West Mains Road, Edinburgh, EH9 3JJ, U.K.
* Corresponding author.
Dr. M. R. Heal
School of Chemistry,
University of Edinburgh,
West Mains Road
Edinburgh
EH9 3JJ UK
Tel: +44 (0)131 6504764
Fax: +44 (0)131 6504743
email: m.heal@ed.ac.uk
Keywords: Passive sampler, Palmes tube, nitrogen dioxide, nitrogen oxides, air quality,
triethanolamine
1
Abstract
A few studies have suggested that the precision and accuracy of measurement of NO2 by
Palmes-type passive diffusion tube (PDT) are affected by the method of preparation of the
triethanolamine (TEA) absorbent coating on the grids. Theses studies have been quite limited
in extent and have tended to evaluate PDT accuracy as zero bias between PDT NO2 value and
the exposure-averaged NO2 determined by co-located chemiluminescence analyser. This
ignores the well-documented intrinsic systematic biases on PDT-derived NO2, such as withintube chemistry and exposure-duration nitrite loss, which may lead to non-zero bias values
irrespective of effects of TEA absorbent preparation method on PDT accuracy. This paper
reports on a statistical analysis of a large dataset comprising 680 duplicated PDT exposures
spanning 146 separate exposures periods, spread over 5 urban exposure locations and a
number of years. In each exposure period, PDTs prepared by between four and six different
grid preparation methods were simultaneously compared. The preparation methods used
combinations of the following: acetone or water as the TEA solvent; 20% or 50% as %TEA in
the solution; and application of TEA solution by dipping grids for several minutes in the
solution before drying and tube assembly, or by pipetting 50 µL of solution directly onto grids
already placed in the PDT cap. These represent the range of preparation procedures typically
used. Accuracy was evaluated as maximised nitrite capture within an exposure. Data were
analysed by General Linear Modelling including examination of interaction between different
aspects of grid preparation method. PDT precision and accuracy were both significantly better,
on average, when the PDT grids were prepared by dipping in TEA solution, and neither
solvent or %TEA used for the dipping solution were important. Where PDT preparation by
pipetting TEA solution onto grids is to be used, better performance was obtained using 20%
TEA in water. A systematic positive bias in PDT measure of NO2, consistent with within-tube
2
oxidation of NO to NO2 and independent of preparation method, was again evident in this
work.
Introduction
Nitrogen dioxide (NO2) is an air pollutant with adverse health associations
1
for which air
quality objectives on ambient air concentration have been set in many countries. It is both
directly emitted and the product of atmospheric oxidation of NO. However, despite increased
controls on some source emissions of NOx (= NO + NO2), levels of NO2 in many locations are
only declining slowly, if at all
2,3
. As a consequence, measurement of ambient NO2 remains a
priority in air pollution management in order to identify where breaches in NO2 objectives
exist and to monitor the efficacy of mitigation action.
The reference method for ambient NO2 measurement is the chemiluminescence analyser.
However, in the UK, as elsewhere, the main approach for indicative assessment of ambient
NO2 is the deployment of Palmes-type passive diffusion tube (PDTs) 4 within extensive Local
Authority networks 2. No agreed standard yet exists for PDT methodology so variations in the
preparation of the triethanolamine (TEA) absorbent and in the post-exposure extraction and
quantification of the captured nitrite from that originally proposed by Palmes et al.
4
have
arisen in practice. Whilst it is accepted that NO2 PDTs are not as accurate or as precise as
continuous analysers, it is also recognised that these methodological differences may
contribute to the observed variations in PDT results in intercomparison procedures 5.
One aspect of the PDT method where there are clearly distinguishable differences in procedure
between laboratories is in the preparation of the TEA-coated grids and a few studies have
3
reported associated differences in the precision and accuracy of the NO2 concentration 6-8. The
three main variations in PDT preparation are that the TEA absorbent may be dissolved in
acetone or water as a solvent, that the concentration of TEA in the solvent may be 50% or 20%
(sometimes 10%), and that the TEA is applied to the stainless steel grids either by submerging
the grids in the TEA solution before removal and drying on absorbent tissue, or by pipetting a
fixed volume (usually 50 μL) of the TEA solution onto two grids already inserted into an end
cap. In one study 7 three preparation methods (dipped 50% TEA in acetone, pipetted 50% TEA
in water, pipetted 20% TEA in water) were compared over three exposures in a dosed tank
environment and reported that tubes prepared by pipetting 50% TEA in water on the grids
yielded lower NO2 than measured by the chemiluminescence analyser and the other PDT
methods. Kirby et al. 6 compared tubes prepared by dipping 50% TEA in acetone or pipetting
10%, 20%, or 50% TEA in water over 10 exposure periods at a single urban background site
and also concluded that tubes prepared by pipetting 50% TEA in water onto grids gave lower
NO2 capture (and were marginally less precise).
Both the above studies are based on a very small number of PDT exposures. Hamilton and
Heal
8
were more comprehensive in their approach, simultaneously trialling eight different
preparation method combinations in 80 exposures spread across three different exposure sites.
They concluded that precision was poorer, on average, for pipetted application methods, and
that pipetted methods were also more variable in accuracy. Preparation methods using 50%
TEA in water led to particularly imprecise data. However, a shortcoming of this latter study
was that pipetting was onto grids in caps into which the tube barrel had already been inserted,
whereas the more general practice is to insert the tube barrel into the cap after the solution has
been pipetted onto the grids.
4
Overall, systematic comparison of different PDT preparation methods remains very limited,
particularly when the pipetting preparation data of Hamilton and Heal 8 are excluded as being
unrepresentative of actual practice. The small numbers of exposure periods and the small
numbers of different preparation methods trialled at any one time also do not allow
investigation of interactions between the preparation variables.
A further important shortcoming of previous work has been equating PDT accuracy with zero
bias between the NO2 concentration derived from the PDT and from a co-located
chemiluminescence analyser. There are good theoretical reasons, supported by observation, to
expect intrinsic biases in PDT measurement, the net magnitude of which will differ with
exposure location and exposure period, irrespective of any inaccuracy associated with PDT
preparation method. Foremost amongst the intrinsic positive biases is within-tube chemical
generation of additional NO2 from co-diffusing NO and O3 9 whose magnitude depends on the
time-varying absolute and relative concentrations of the three species during the exposure 10,11.
This will be affected, amongst other factors, by distance of a particular PDT exposure location
to strong sources of NO, e.g. busy roads. Secondly, positive bias due to shortening of the
diffusion path length by wind-induced turbulence at the mouth of the tube
12
, if significant, is
also likely to vary according to the exposure location and period. Similarly, intrinsic negative
bias caused by exposure-duration dependent loss of captured nitrite may also vary with
exposure location and time of year 11. The net effect of these factors will vary according to the
individual exposure environment so it is erroneous to use tendency to zero bias as the metric
for PDT accuracy, particularly when comparing results from different PDT exposures.
The work reported here overcomes these issues. It is a statistical analysis of PDT precision and
accuracy of a large dataset of 146 separate PDT exposures in which PDTs prepared by at least
5
four out of seven different grid preparation methods were simultaneously exposed in duplicate
during each exposure. The data are derived from a range of urban exposure locations, over a
period of several years, and includes the work of several analysts. Because several
combinations of preparation methods were trialled every exposure, it is possible to test for
statistical interaction between different aspects of PDT preparation method; for example, does
effect of solvent differ according to % solvent. Also, the assessment of PDT accuracy used
here is independent of any exposure-related intrinsic bias in PDT NO2. The dataset includes
the dipped (but not pipetted) methods data previously reported in Hamilton and Heal 8 but the
other 732 PDT NO2 data are unpublished.
Methods
The acrylic diffusion tubes, polyethylene caps and stainless steel grids used for the PDT
exposures described here were obtained from Gradko International. All components were reused (subject to visual inspection of their continued integrity), but were cleaned thoroughly
between use with detergent solution and deionised water. Tube physical dimensions (mean 
95% CI from measurement of a sub-sample) were length 7.1  0.1 cm, internal cross-section
area 0.92  0.02 cm2.
The dataset of PDT NO2 measurements comprises PDT exposures at five different sites in
Edinburgh. Three are classified as roadside (Haymarket, Castle St. and Queen’s St.), one as
urban centre (the former AURN site at Princes St. Garden) and one as urban background (the
relocated AURN site at St. Leonard’s). The median (min-max) exposure-averaged NO2
concentrations across all sites, as measured by co-located chemiluminescence analysers, was
39.3 (13.2 – 86.5) µg m-3. The exposures were spread intermittently over a period of more than
5 years between November 2001 and February 2007. Over this period four different analysts
6
prepared and analysed the PDTs, always following exactly the same written protocols. The
fact that different analysts have contributed to the dataset presented here emulates the realworld situation in which different laboratories and analysts contribute to monitoring network
data.
PDT preparation methods under trial comprised different combinations of the two different
“levels” of each of the following three preparation method “factors”:
(1) Factor “Application Method,” with levels “dipped” (grids coated with TEA by soaking
them for 10 min in the appropriate solution of TEA followed by drying of the grids on tissue
and tube assembly with two grids per tube) or “pipetted” (grids coated by pipetting 50 μL of
the TEA solution onto two stainless steel grids already placed in a cap followed by completion
of tube assembly);
(2) Factor “Solvent,” with levels “acetone” or “water”;
(3) Factor “%TEA” (in solvent), with levels “20%” or “50%”.
For the methods in which grids were prepared by pipetting, the solution of TEA was spread as
evenly as possible around the grid surface with a spreading action of the pipette tip.
Duplicates of PDT preparations were deployed in every PDT exposure. These provide %RSD
(relative standard deviation) data for evaluation of precision. Although all PDT exposures
were co-located with a chemiluminescence analyser, for the reasons given in the introduction
%BIAS of PDT with respect to the analyser is not used here for evaluation of PDT accuracy.
Instead, the following measure of accuracy was adopted. For any given exposure location and
exposure period all co-located PDTs should be subject to the same intrinsic biases of withintube generation of NO2, wind-induced shortening of diffusion path length, and exposureduration reduction in nitrite. Since the goal of diffusion tube methodology is to attain
7
stoichiometric trapping of NO2 arriving at the absorbent into nitrite then, other things being
equal, greater PDT-derived NO2 concentration in a given exposure equals greater PDT
accuracy. (PDT-derived NO2 concentration is a direct proxy for nitrite captured). The method
of preparing grids by dipping in solutions of 50% TEA in acetone was trialled in every
exposure in this dataset. Therefore accuracy of PDT preparation method is evaluated using the
CAPTURE metric, where CAPTURE = %[(duplicate mean PDT NO2 for prep. method X)/
(duplicate mean PDT NO2 for prep. method dipping, acetone, 50% TEA)]. Lower values, on
average, of CAPTURE is indicative of poorer accuracy, on average.
Regardless of their method of preparation, all PDTs were subsequently handled equivalently
and analysed in the same way to quantify the nitrite trapped in the absorbent during exposure.
The nitrite was first extracted into 1.5 mL of deionised water in-situ in the tube, and then 1.65
mL of acidified mixed sulphanilamide:NEDA solution (in reagent mass ratio 1:0.007) added to
form the diazo chromophore. Absorbance intensity was measured at 540 nm in a dual beam
UV/vis spectrometer using solution from an equivalently-treated blank tube (i.e. reagents
present but zero nitrite) in the reference beam. Independent duplicate sets of nitrite calibration
standards were prepared for each analysis. The average ambient NO2 concentration during the
exposure was calculated from the nitrite calibration graphs using 0.154 cm2 s-1 as the diffusion
coefficient of NO2 in air. This is the value of diffusion coefficient originally recommended by
Palmes et al. 4 and used for many years throughout the UK national and local authority NO2
PDT networks. Recently the UK Working Group on NO2 PDT harmonisation, referring to
work by Massman 13, has recommended 0.146 cm2 s-1 as the value for the diffusion coefficient
appropriate to a UK average ambient temperature of 284 K 5. However, because it is necessary
to report measured concentrations of NO2 in mass units at the EU standard reporting
temperature of 293 K (particularly when comparing PDT data to chemiluminescence data
8
which are likewise reported at a temperature of 293 K), the NO2 mass concentration derived
using this diffusion coefficient must be subsequently temperature corrected by a factor
284/293 = 0.969 5. The net effect of both these corrections can be achieved in one step using a
diffusion coefficient value of 0.151 cm2 s-1. Application of the updated specification for the
diffusion coefficient would increase each value of PDT-derived NO2 concentration used in this
paper by ~2%, but would have no effect on the evaluation of PDT precision and accuracy via
%RSD and CAPTURE since both are relative values.
The total dataset available comprises 680 duplicated PDT exposures spanning 146 separate
exposure locations and periods. There are almost no missing data within this total: 7 missing
%RSD values because of loss of one or more replicate tube, and 3 missing CAPTURE values
where both replicates are missing. The dataset has not been “cleaned” of very poor %RSD
values, all of which remain included.
Of the eight possible combinations of tube preparation factors and levels, the preparation
method of pipetting 20% TEA in acetone was not trialled since this method has not been
reported in the literature nor used by any of the labs in the UK PDT networks. For the seven
other combinations of tube preparation factors and levels under investigation, not all were
compared in every exposure period, since this would have involved an excessive number of
parallel deployments, but each exposure always consisted of a minimum of four, and up to six,
of the possible combinations. (Average number of different preparation methods trialled per
exposure period = 680/146 = 4.7). There is an imbalance in the dataset in number of exposures
associated with each possible preparation combination, ranging from 146 exposures for each
of the two preparation methods 50% TEA in acetone, dipped, and 50% TEA in water, dipped,
to 30 exposures for the preparation method 20% TEA in water, pipetted. Importantly,
9
however, all seven tested preparation methods were trialled by at least two of the four analysts
which reduces the possibility that any observed differences in NO2 measurement between
differently-prepared PDTs are driven solely by an analyst-associated cause.
All PDT exposures were 1-week. While this does not emulate the majority practice of UK
networks, which is for 1-month exposures, it is not considered to be relevant to the data
analysis, as both %RSD and CAPTURE are expressed relative to other PDT values in the
same exposure. In fact, an advantage of 1-week exposures is that the confounding effect of
decreasing PDT NO2 values with exposure duration, which may well be variable, should be
minimised 11.
General linear modelling (GLM) in Minitab v.14 was used to test for significant factors acting
on the two dependent variables %RSD and CAPTURE being investigated. The full GLM
comprised: (a) the three factors “Application Method”, “Solvent” and “%TEA” as the maineffect fixed factors under test; (b) the three pairwise combinations of these factors to test for
significant interactions; (c) the two factors “Analyst” and “Exposure site” as additional maineffect random factors i.e. factors not under specific experimental test, but which might also
have influence on the dependent variables. There were insufficient combinations of trials to
examine for interaction between “Analyst” and the three main factors under test. The extent of
interaction between the three main test factors was examined closely since it is important to
check whether a non-significant main effect of a factor results from two opposing trends of a
second factor in combination with the first factor.
Regardless of the significance values of the two random factors “Analyst” and “Exposure
site”, the GLMs were repeated with one or both of these factor designations removed from the
10
model in order to examine whether this led to any difference in the interpretation of the effects
of the fixed factors on the dependent variables.
Results
Evaluation of preparation method on precision
The mean %RSD values for all exposures of each preparation technique investigated are
illustrated in Figure 1. The p values associated with terms in the GLM of the total %RSD
dataset are given in Table 1. The GLM confirms the expectation that PDT precision is
independent of the exposure location, so Figure 2 shows the main and interaction effects plots
for the model in which exposure site is not a specified factor.
Figure 1 shows that PDTs prepared by dipping grids in solution are more precise, on average,
than PDTs prepared by pipetting the solution onto grids in caps. The high significance
associated with this observation is confirmed in Table 1 and Figure 2. In contrast, there is no
evidence of significant differences in PDT precision for either acetone or water as the solvent
or for either 20% or 50% TEA composition. Analyst is a significant factor which reflects the
reality that different analysts differ in the quality of their precision. However, the difference in
precision between analyst is smaller than the difference in precision associated with
application method. Since different analysts trialled different combinations of preparation
method factors and levels, the significant analyst term should not unduly influence the
interpretations of the effects of the fixed factors under test.
The above interpretations of the effects of the main factors on precision are not compromised
by the presence of significant interactions between the main factors. There is some trend
towards better dipped method precision with 20% TEA in solvent and better pipetted method
11
precision with 50% TEA in solvent but the effect is small compared with the difference in
precision between dipped and pipetted methods.
Evaluation of preparation method on accuracy
The mean CAPTURE values for each preparation method are illustrated in Figure 3 and the pvalues derived from application of the GLMs to the data are summarised in Table 2. These
show significantly greater nitrite capture, taken to represent greater accuracy, on average, for
PDTs in which grids are prepared by dipping rather than by pipetting. Dipped method PDT
NO2 values are also more significantly correlated with the corresponding chemiluminescence
analyser data than those from the pipetted methods (values not reported), a metric which is
also interpreted as indicating better performance of a PDT method (although still subject to
distortion because of different intrinsic PDT biases in different exposure situations).
Exposure location is not a significant factor which supports the rationale for use of the
CAPTURE metric for evaluation of accuracy. (Exposure location is a significant factor in a
GLM using %bias as the dependent variable). Analyst is a significant factor. This cannot be
due to systematic differences between analysts in the nitrite extraction and calibration part of
the NO2 determination since any such error would influence all nitrite analyses for a given
exposure period equally. Instead the significant influence of analyst is a consequence of the
different distributions between analysts of preparation methods trialled. This is not relevant to
a statistical analysis of CAPTURE values so the main and interaction effects plots shown in
Figure 4 are for the model excluding site and analyst as factors.
The data show that, as for precision, accuracy is not influenced by the solvent. Accuracy is
significantly influenced by %TEA but this needs to be interpreted in the context of significant
12
interaction: greater accuracy is associated with 20% TEA solutions when coating grids by
pipetting but accuracy is not influenced by %TEA when coating grids by dipping.
Discussion
A consistent picture emerges from the statistical evaluation of both precision and accuracy.
Dipped preparation methods provide both better precision and greater nitrite capture
(interpreted here as greater accuracy), on average, than do pipetting preparation methods.
Neither precision or nitrite capture is influenced by use of either acetone or water as the
solvent. There is a significant trend for 20% TEA composition to yield greater nitrite capture
than 50% TEA composition in pipetting methods, but %TEA is not an important factor for
nitrite capture in dipping methods nor for precision for any method trialled. The observation of
lower nitrite capture when pipetting with higher %TEA solution is consistent with the earlier
smaller trials 6,7.
The conclusions from this study are supported by a recent analysis by Air Quality Consultants
Ltd. of data from PDT exposures co-located with chemiluminescence analysers accumulated
from local authorities around the UK
14
. These authors applied five different quantitative and
semi-quantitative measures of PDT precision and accuracy to a dataset comprising 161 annual
co-location studies involving 21 laboratories to try and tease out which of several variations in
PDT preparation and analysis methodology used by the laboratories led to better overall PDT
performance. In respect of issues associated with variations in PDT preparation method the
following were concluded (using Laxen et al. 14 phrasing): some evidence that dipping of grids
provides a better performance than tubes prepared by pipetting onto grids; a clear pattern that
tubes prepared with grids soaked for 10 min or more performed better than tubes prepared
with grids dipped for less than 1 min; a clear pattern that tubes prepared with grids that have
13
been dried before assembly perform better than those with grids that are wet when the tube is
assembled; and a clear pattern that tubes prepared using 20% TEA in water perform better
than those using 50% TEA in acetone. A caveat on the last conclusion is that the dataset did
not include preparation methods with 20% TEA in acetone or 50% TEA in water. Also, these
workers’ analysis could not examine for interaction between %TEA, solvent and grid
application technique.
It is to be expected that poorer nitrite capture efficiency is associated with the same method(s)
that give poorer precision. Given there is a physical upper limit to the amount of nitrite that
can be captured, then greater variability in nitrite capture must lead to lower capture
efficiency, on average.
The poorer performance of pipetting methods revealed in this and the Laxen et al.
14
analyses
is likely driven, at least in part, by variability associated with different analysts’ approaches to
execution of the pipetting technique, as compared with the likely smaller inter-analyst
variability associated with execution of the dipping technique. The latter is essentially analyst
independent, i.e. TEA coverage via surface tension followed by drying. Thus, for a single
lab/analyst, it may well be possible to obtain pipetted method precision to match dipped
method precision. Even so, Gerboles et al. 15 reported substantial sensitivity of PDT precision
to whether a pipetted solution (of 10% TEA in water) was spread all across the grid surface
compared with the same volume pipetted as a single drop in the centre of the grids.
In addition to greater variability in analyst technique, pipetting methods may lead to poorer
precision and nitrite capture because of greater variability in both amount and coverage of
TEA coating caused, perhaps, by less even coverage of grids from pipette dispensing, less
14
absolute amount of TEA added to grids in pipetted solution compared with coating by dipping,
loss of viscous TEA sticking to pipette tip, and handling of assembled tubes before the grids
are thoroughly dry leading to dribbling of TEA solution down tube walls. Pipetting solutions
of TEA in acetone will be particularly subject to variability because of the difficulty of
consistently drawing up and dispensing a solution comprised of such a volatile solvent.
Although not the primary focus of the study, it is relevant to report that the PDT-derived NO2
concentrations in this work were larger, on average, than the corresponding one-week
exposure-averaged NO2 measured by chemiluminescence analyser (although, as has been
demonstrated, the extent of positive bias varied with different PDT preparation method). For
example, median bias across all sites for PDTs prepared by dipping grids in 50% acetone was
26% (n = 140, bias values not available for 6 exposures because of missing analyser data). The
median bias for this method for individual sites varied between 17% (n = 52) at the Haymarket
site to 57% (n = 22) at the Queen’s Street site. For comparison, the median exposure-averaged
analyser NOx/NO2 ratio was 2.1 across all sites, and 2.2 and 2.0 for the Haymarket and
Queen’s Street sites, respectively. The data therefore provide further evidence of intrinsic
over-read of NO2 by PDT consistent with within-tube oxidation of a proportion of the NO also
present at the exposure location10,11,16, and confirms the problem of using zero bias as a
measure of accuracy.
Conclusions
Accumulated evidence clearly points towards greater consistency, i.e. method robustness, on
average, in PDT precision and accuracy when the PDT absorbent grids are prepared by
thoroughly soaking in TEA solution and subsequent drying before tube assembly. The
proportion TEA and solvent used for the solution are not important, although it is strongly
15
recommended that only a single solvent-%TEA combination is specified in preparation
protocols by way of harmonisation and standardisation. Where PDT preparation by pipetting
TEA solution onto grids pre-installed into caps is retained as a method, it is strongly
recommended that this should be a standardised volume (e.g. 50 µl) of a solution of 20% TEA
in water. The findings from this work support the instructions for permitted PDT absorbent
grid preparation methods recently issued to UK laboratories 5.
A systematic positive bias in PDT measure of NO2, consistent with within-tube oxidation of
NO to NO2 and independent of preparation method, is again evident in this work.
The above conclusions are independent of any variation in PDT NO2 measurement that may
be caused by variation in the procedure used for extraction and quantification of trapped nitrite
after PDT exposure. It is recommended that this aspect of NO2 PDT methodology also be
evaluated.
Acknowledgements
The assistance in laboratory work over a number of years from students in the School of
Chemistry, University of Edinburgh, is gratefully acknowledged, as are Janet Brown and
colleagues from the City of Edinburgh Council for access to exposure locations and provision
of chemiluminescence analyser data.
16
References
1. WHO, Health aspects of air pollution with particulate matter, ozone and nitrogen
dioxide. Report on a WHO Working Group, EUR/03/5042688, World Health
Organisation, 2003.
2. AQEG, Nitrogen Dioxide in the United Kingdom, Air Quality Expert Group, UK
Department for Environment, Food and Rural Affairs, PB 9025, London, 2004.
3. AQEG, Primary Nitrogen Dioxide in the United Kingdom. Fourth report of the Air
Quality Expert Group, UK Department for Environment, Food and Rural Affairs,
PB12779, London, 2007.
4. E. D. Palmes, A. F. Gunnison, J. DiMattio and C. Tomczyk, American Industrial
Hygiene Association Journal, 1976, 37, 570-577.
5. Defra WG, Diffusion tubes for ambient NO2 monitoring: practical guidance for
laboratories and users. A report by the Defra Working Group on Harmonisation of
Diffusion Tube Methods, AEA Energy & Environment, Didcot, UK. Report no.
AEAT/ENV/R/2504.
www.airquality.co.uk/archive/reports/cat05/0802141004_NO2_WG_PracticalGuidance_
Issue1a.pdf, 2008.
6. C. Kirby, M. Fox and J. Waterhouse, J. Environ. Monitor., 2000, 2, 307-312.
7. A. Loader, Investigation of the effects of preparation technique on performance of
nitrogen dioxide diffusion tubes, Report no. AEAT/ENV/R/0563, Harwell, UK, 2001.
8. R. P. Hamilton and M. R. Heal, J. Environ. Monitor., 2004, 6, 12-17.
9. M. R. Heal and J. N. Cape, Atmos. Environ., 1997, 31, 1911-1923.
10. M. R. Heal, M. A. O'Donoghue and J. N. Cape, Atmos. Environ., 1999, 33, 513-524.
11. M. R. Heal, C. Kirby and J. N. Cape, Environ. Monitor. Assess., 2000, 62, 39-54.
12. A. J. Gair and S. A. Penkett, Atmos. Environ., 1994, 29, -.
13. W. J. Massman, Atmos. Environ., 1998, 32, 1111-1127.
14. D. H. P. Laxen, B. Marner and S. Donovan, Analysis of factors influencing diffusion tube
performance. A report prepared for Defra and the Devolved Administrations by Air
Quality Consultants Ltd., Report no. 504/2/F1,
http://www.airquality.co.uk/archive/reports/cat05/0807101007_Factors_Influencing_Dif
f_Tube_Performance.pdf, Defra, UK, 2008.
15. M. Gerboles, D. Buzica and L. Amantini, Atmos. Environ., 2005, 39, 2579-2592.
16. C. Kirby, M. Fox, J. Waterhouse and T. Drye, J. Environ. Monitor., 2001, 3, 150-158.
17
Tables
Table 1: GLM p-values associated with the effects of named factors and factor interactions on
PDT precision as assessed by the %RSD values for duplicate exposures. Terms significant at
the p < 0.05 level are highlighted in bold.
Term
Analyst
Exposure site
Application
Solvent
%TEA
Application*Solvent
Application*%TEA
Solvent*%TEA
Full model
Model excluding
site as factor
0.006
0.331
0.000
0.152
0.905
0.658
0.104
0.403
0.015
0.000
0.151
0.904
0.654
0.103
0.403
Model excluding
site & analyst as
factors
0.000
0.196
0.579
0.637
0.037
0.310
Table 2: GLM p-values associated with the effects of named factors and factor interactions on
PDT accuracy as assessed by the CAPTURE values (the PDT NO2 value relative to the NO2
derived from the PDT prepared by dipping, 50% TEA in acetone, in the same exposure).
Terms significant at the p < 0.05 level are highlighted in bold.
Term
Analyst
Exposure site
Application
Solvent
%TEA
Application*Solvent
Application*%TEA
Solvent*%TEA
Full model
Model excluding
site as factor
0.000
0.141
0.000
0.971
0.000
0.274
0.000
0.000
0.000
0.000
0.971
0.000
0.274
0.000
0.000
18
Model excluding
site & analyst as
factors
0.000
0.511
0.000
0.113
0.000
0.000
Figure Captions
Figure 1: Mean ( 1 std error) of duplicate %RSD for all PDT exposures of each preparation
method investigated. The number of %RSD data contributing to each mean value is given
above each bar.
Figure 2: Results of application of a general linear model to PDT precision data (%RSD
values) that incorporates application method, solvent and %TEA as fixed-effect factors,
analyst as a random-effect factor and the pairwise interactions of the three fixed-effect factors.
Main effects are illustrated in the upper panels and interactions in the lower panels.
Figure 3: Mean ( 1 std error) of the CAPTURE metric for all PDT exposures of each
preparation method investigated. Alls values of CAPTURE for the PDT preparation method
dipping in 50% TEA in acetone are 100% by definition. The number of CAPTURE data
contributing to each mean value is given above each bar.
Figure 4: Results of application of a general linear model to PDT accuracy data (represented
here by the CAPTURE metric) that incorporates application method, solvent and %TEA as
fixed-effect factors and the pairwise interactions of the three fixed-effect factors. Main effects
are illustrated in the upper panels and interactions in the lower panels.
19
Figure 1
18
n = 30
16
14
Mean %RSD
n = 100
12
n = 66
n = 145 n = 143
10
n = 80
n = 109
8
6
4
2
0
dip,
ace,
50%
dip,
water,
50%
dip,
ace,
20%
dip,
water,
20%
20
pip,
ace,
50%
pip,
water,
50%
pip,
ace,
20%
pip,
water,
20%
Figure 2
Mean %RSD
application
solvent
13
12
11
10
9
8
dipped
pipetted
acetone
water
%TEA
Analyst
13
12
11
10
9
8
20%
50%
acetone
w ater
A
20%
C
D
50%
application
solvent
%TEA
21
B
13
12
11
10
9
8
application
dipped
pipetted
13
12
11
10
9
8
solvent
acetone
water
Mean CAPTURE wrt dip, ace, 50%
Figure 3
120
100
n = 143 n = 80
n = 146
n = 110
n = 30
n = 102
80
n = 66
60
40
20
0
dip,
dip,
ace, water,
50% 50%
dip,
dip,
ace, water,
20% 20%
22
pip,
pip,
ace, water,
50% 50%
pip,
pip,
ace, water,
20% 20%
Mean CAPTURE wrt dip, acetone, 50% TEA
Figure 4
Application
Solvent
105
100
95
90
85
dipped
pipetted
acetone
water
%TEA
105
100
95
90
85
20%
50%
acetone
w ater
20%
50%
110
100
Application
dipped
pipetted
90
Application
80
110
100
Solvent
90
80
%TEA
23
Solvent
acetone
water